An Experimental Study and Flight Testing of Active Aeroelastic Aircraft Wing Structures

Transcription

1 An Experimental Study and Flight Testing of Active Aeroelastic Aircraft Wing Structures Joana Luiz Torres da Rocha Aeronautical Engineering Degree, Portuguese Air Force Academy, 2001 A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of in the Department of Mechanical Engineering University of Victoria A11 rights reserved. This thesis may not be reproduced in whole or in part, by photocopy or other means, without the permission of the author.

2 Supervisors: Dr. Afzal Suleman Abstract An experimental investigation on active control of aeroelastic aircraft wing structures using piezoelectric actuators and sensors is presented. To this end, wind tunnel and remotely piloted vehicle wing models were designed, fabricated, installed, and tested. Computational structural and aerodynamic wing models were created, in order to determine the wing natural frequencies and modal shapes, and to predict the flutter speed. A digital controller was designed and implemented. Open and closedloop vibration and flutter tests were conducted in the wind tunnel and in flight, with excellent correlation achieved with computational predictions. Two different active wing concepts were analyzed: the first model consists of a wing with piezoelectric actuators attached to the wing skin, and the second wing model has piezoelectric actuators mounted in the main spar. The experimental results obtained have shown that the adaptive wing response had improvements in almost all the RPV flying conditions compared to the corresponding passive wing vibration, for both the active skin and the active spar wing concepts. Also, it was demonstrated that the flutter speed of the active wings increased compared to the corresponding passive wings.

3 Table of Contents Abstract List of Tables List of Figures vii 1 Introduction 1.1 Motivation Background Overview of the Active Aeroelastic Structures Developments State of the Art in Multifunctional Materials Structure of the Thesis... 2 Computational Analyses 2.1 Fundamentals of Flutter Finite Element Analysis Adaptive Skin Wing Adaptive Spar Wing Flutter Analysis Adaptive Skin Wing Adaptive Spar Wing... 3 Wind Tunnel Tests 3.1 Experimental Apparatus Wind Tunnel Tests Articles The Controller Electronic Equipment Tests Objectives and Procedures Adaptive Skin Wing Tests...

4 TABLE OF CONTENTS iv Vibration Tests Damping Analysis Flutter Analysis Adaptive Spar Wing Tests VibrationTests Damping Analysis Flutter Analysis Flight Tests Experimental Apparatus The RPV Additional Electronic Equipment Tests Objectives and Procedures Adaptive Spar Wing Tests Conclusions and Future Work 90

5 List of Tables 2.1 Adaptive Skin wing first ten natural frequencies Adaptive Spar wing first five natural frequencies Adaptive Skin wing flutter results, using the g-method Wing material properties Average and maximum displacement values for the passive and active skin wing configurations Displacements improvements of the active skin wing compared with the passive skin wing Damping of the active skin wing due to tail vibration, tested at 10, and 37.5mls Damping of the wing for the seventh natural mode. tested at and 37.5mls Average and maximum displacement values for the passive and active spar wing configurations Displacements improvements of the active spar wing compared with the passive spar wing Average and maximum displacement values for the passive and active spar wing configurations. using the RPV without the flexible tail Displacements improvements of the active wing compared with the passive wing. using the RPV without the flexible tail Damping of the active spar wing due to tail vibration. tested at and 30mls Damping of the wing for the first torsion mode. tested at and 30mls RPV external dimensions RPV areas RPV weights and loadings RPV performance data...

6 LIST OF TABLES vi 4.5 Average and maximum displacement values for the passive and active spar wing configurations, in the flight tests Displacements improvements of the active spar wing compared with the passive spar wing, in the flight tests

7 vii List of Figures Illustration of a piezoelectric actuator deformation Force-deflection output of a typical piezoelectric actuator (a) Diagram of a piezoceramic stack; (b) Typical bimorph bender actuator ACX Quickpack Actuator Single degree of freedom system with translational mass Aeroelastic functional diagram Adaptive Skin wing finite element mesh. defined in the ANSYS program. 27 Adaptive Skin wing first ten mode shapes Adaptive Spar wing finite element mesh. defined in the ANSYS program. 30 Adaptive Spar wing first five mode shapes.... Adaptive Skin wing V-f graphic.... (a) Adaptive Skin wing V-g graphic; (b) V-g graphic zoom near the zero damping.... Adaptive Skin wing flutter modes shapes.... Photograph of the Air Force Academy Aeronautical Laboratory wind tunnel... Photograph of the RPV model with active wing mounted in the wind tunnel section... Photograph of the RPV wind tunnel model fuselage... Schematic view of the interconnection between the RPV fuselage and wing (not to scale)... Schematic view of the wing airfoil shape and components... Photograph of active skin wing internal assemblage (without piezo- electric~)... Photograph of active skin wing with piezoelectric sensors and actuators mounted in the wing lower surface....

8 LIST OF FIGURES... Vlll 3.8 Schematic view of the piezoelectric sensor (small patch) and actuators (big patches) placement inside the active skin (gray panel) wing Photograph of the RPV model with active skin wing in the wind tunnel Scheme of the active skin wing control vibration process (the rectangular patches represent the piezoelectric actuators and the gray panels the carbon fibre plates) Schematic view of the wing airfoil shape and hollow squared beam Photograph of active spar wing internal assemblage (without piezoelectric~) Photograph of active spar wing with piezoelectric sensors and actuators Schematic view of the piezoelectric sensor (small patch) and actuators (big patches) placement in the active spar (gray beam) wing Lead-zirconate-titanate piezoelectric sensors ACX Quickpack 40W actuator characteristics Photograph of the dspace MicroAutoBox 1401/ Photograph of the SA-10 power amplifier Photograph of the signal conditioning circuit and SA-10 amplifier box Photograph of the 16.8V (left) and 14.4V (right) batteries Scheme of the complete hardware system for the wind tunnel and flight tests Block diagram of conceptual active wing control model Diagram representing the control model implemented for wind tunnel tests Average displacements of both passive and active skin wing configurations, in the wind tunnel tests Maximum displacements of both passive and active skin wing configurations, in the wind tunnel tests Damping calculation scheme Damping curves of both passive and active skin wing configurations, due to the tail vibration Curves of the wing damping for the seventh natural mode at 71.9Hx, and polynomial extrapolation to zero damping Average displacements of both passive and active spar wing configurations, in the wind tunnel tests Maximum displacements of both passive and active spar wing configurations, in the wind tunnel tests Average displacements of both passive and active spar wing configurations, in the wind tunnel tests (RPV without tail) Maximum displacements of both passive and active spar wing configurations, in the wind tunnel tests (RPV without tail)

9 LIST OF FIGURES 3.33 Damping curves of both passive and active spar wing configurations, due to the tail vibration Curves of the wing damping for the first torsion mode, and extrapolation polynomial to zero damping Picture of the RPV model ready to the flight tests Photograph of the telemetry airbone station components Photograph of the telemetry ground station components Picture of the real-time data display Photograph of the RPV fuselage containing all the flight tests equipment. 4.6 Diagram representing the control model implemented for flight tests. 4.7 Amplitude (V/5) of the wing versus time, in speed window

10 Acknowledgements During the development of this research I had the precious support of some people that made this work possible and enjoyable. Because of this, I would like to thank all of you very much. To Dr. Afzal Suleman for giving me the opportunity to be past of his research group, for giving me support and motivation during all the phases of the study. To LtCol. Ant6nio Pedro Costa, for giving me technical and logistic support, for his example and confidence. To Engineer Paulo Moniz, for helping me in the stage of the wind tunnel tests, for the conceptual design of the RPV, for his helpful comments and discussions. To Capt. Costa, for the assistance in the fabrication of all the wind tunnel and flight models, and for being the pilot during the flight tests. To Bruno Carreiro, for the flight tests telemetry system setup. To the Portuguese Academy Aeronautical Laboratory personnel: Captains Dores and Madruga, Lieutenants Silva and Pinheiro, to Sargeants Ramos and Fernando for their support in the installation of the electronic equipment necessary for wind tunnel and flight tests, to Sargeants Fernandes for the support in the computers setup, Privates Costa, Brand50 and Filipe, and to D. Fernanda. To all members of our research group, for sharing your enthusiasm with me: Dr. Suleman, Sandra Makosinski, Luis Falc50, Diogo Santos, David Cruz, Scott Burnpus, Gon~alo Pedro, Marc Secanell, Ernest Ng and Ahmad Kermani. Finally and very important, I want to thank to my family: to Bruno, for being always there, to Helena, for giving me her genuine and inner encouragement; to my parents, Luis and Maria da Luz, and sisters, Ink and Libiinia, for their example and emotional encouragement. All of you made it possible.

11 To Helena and Bruno

12 Chapter 1 Introduction When an aircraft is flying, the aerodynamic forces cause deformations in the structure (especially in the wings) during the entire flight envelope. These deformations are known as vibrations defined as a motion that repeats itself in time. Although these vibrations are necessary and inevitable, they are also responsible for structural damage. This damage can occur in two different ways: abruptly or caused by fatigue. The fatigue damage is caused by the continuous low vibration of the structure. Abrupt damage happens when a catastrophic aeroelastic event takes place, for example when the aircraft experiences wing divergence or flutter. In both cases, the damage can be catastrophic and cause the loss of people and aircraft. As a result, an important issue in aircraft design is the study of the aeroelastic response of the flight vehicle. The work presented in this thesis concerns the active aeroelastic response of aircraft structures. In particular, the study of a Remote Piloted Vehicle (RPV) wing deformations and the reduction of these vibrations using piezoelectric actuators and sensors was performed, in order to increase the flight envelope in terms of flutter. It is shown that the piezoelectric sensors and actuators are effective when used in small scale flight vehicles and a considerable increase in flutter speed was observed.

13 CHAPTER 1. INTRODUCTION 2 Two adaptive wing concepts are proposed in this thesis: a wing with piezoelectric materials mounted in the wing surface (adaptive skin concept), and a wing with piezoelectric materials mounted in the main spar (adaptive spar concept). The research was carried out in three logical stages: first, computational analyses were performed to predict the response of the adaptive wings in passive mode; next, wind tunnel tests were carried out to validate the computational models; and finally a flight test was performed to verify the performance in real flight conditions. The computational study was performed using commercial finite element (ANSYS) and aeroelastic analysis (ZAERO) programs. The ANSYS program was used to determine the wing natural frequencies and modal shapes of the adaptive wings in passive mode. ZAERO program calculates the wing flutter speed. In wind tunnel and flight tests, the wing was tested in two different configurations: with and without the actuation of piezoelectric materials, i.e., in the active and passive modes. In other words, the wing with vibration control and the wing in free vibration (i.e., without vibration control). After obtaining the vibration results of both wing configurations, it was possible to analyze the differences between them, and measure the wing vibration improvements, i.e., the reduction of vibration in terms of average cycles and magnitude. For control and data acquisition, the MATLAB program and DSPACE tools were used. In the next Section, the motivation of this thesis is described in detail. The background is presented next in Section 1.2. An overview of the past studies and developments in the area of active aeroelastic structures is presented in The state of the art in multifunctional materials is presented in Finally Section 1.3 descrives the content of the various Chapters in the thesis.

14 CHAPTER 1. INTRODUCTION Motivation A RPV is the predecessor of an Unmanned Aerial Vehicle (UAV). The main difference between an RPV and UAV is that the RPV is not a self-piloted aircraft. The RPV needs to have someone flying it, using remote control. Because of that, RPVs still have a range problem, which is limited by the radio transmitter range. The UAV is self-piloted, i.e., autonomous, and carries a computer with the entire flight envelope previously programmed. They can carry cameras, sensors, communication equipment or other payloads. Therefore, they can be used in reconnaissance, intelligence-gathering role and combat missions. Nowadays, UAVs can be divided in two categories: Tactical and Endurance (long range) [I]. Most importantly, UAVs are today widely used in military reconnaissance and forest fire observation missions. In the last two decades, the technological developments in the areas of materials and computer sciences have been very promising. The combination of multifunctional materials with faster computers and data acquisition systems has resulted in adaptive systems. The development of materials science made the materials multifunctionality possible, such as piezoelectricity. On the other hand, the development of computational sciences made advances in areas such as design, manufacture and control possible. An adaptive system is a structure with embedded sensors, that provide information about its environment, for instance, forces, tension field, displacements, etc. Then, this data is used by a processor and a control module in order to generate a response to the actuators, attached to the structure, in order to change the structure properties. The multifunctional materials applied to structures can mitigate structural problems involving vibration suppression, noise reduction and shape control.

15 CHAPTER 1. INTRODUCTION 4 Adaptive systems are also called "smart structures". These structures are known as "smart" because they sense changes in their environment and respond accordingly to these changes [2]. In the past, some passive solutions were used to solve aeroelastic dynamic problems, such as increasing the structural rigidity or balancing the mass. Increasing the structural rigidity makes the structure heavier. Here, the use of active control systems using distributed actuation is proposed and this approach can result in an improved structural response without the added weight penalty. The most popular multifunctional materials are the piezoelectrics, electrostrictives, magnetostrict ives, shape memory alloys, electrorheological and magnetorheological fluids. Multifunctional materials are also known as "smart materials". The "smart structures", mentioned in the last paragraph, integrate "smart materials" and controllers. These materials respond to external stimuli like electric, magnetic or thermal fields. In particular, piezoelectric materials can operate as both sensors and actuators. In sensor mode, they produce voltage when a mechanical strain is applied. In actuator mode, they undergo elongation when an electric field is applied [3]. In general, piezoelectric materials are more suitable for operation at high frequencies compared with t he other multifunctional materials. However, since they are easily breakable, the manufacture and handling of piezoelectric crystals are difficult. Although the ceramic properties of the piezoelectrics are enough for several applications, when large displacements and forces are intended or certain frequency ranges are expected, the use of other type of materials is necessary.

16 CHAPTER 1. INTRODUCTION 1.2 Background Overview of the Active Aeroelastic Structures Developments Aeroelasticity is the interaction between elastic, inertial and aerodynamic loads, acting on the aircraft in operating conditions. In normal flight conditions, these loads may cause the aircraft to become unstable. In real life, aeroelastic events can be static or dynamic phenomena. A classical example of a static problem is the divergence phenomenon, and flutter is possibly the most important dynamic event in aeroelasticity. As an illustration of this event importance, the flutter envelope prediction is crucial to the certification of civil and military aircrafts. Also, the active suppression of aeroelastic instabilities such as flutter or divergence leads to improved performance. Threfore, many control strategies have been applied to suppress flutter or control unacceptable wing motion. Concerns and considerations about aeroelasticity were considered very early in the history of aviation. The failure of the Langleys monoplane, in October 1903, was considered to be caused by aeroelastic problems, possibly by the wing torsional divergence [4]. The Wright brothers took advantage of the wing flexibility to control what is known as the first successful flight, in December Instead of ailerons or flaps to control their airplane, they twisted the craft wings as a mean to control its rolling motion. This system avoided the extra weight of the aileron control surfaces ~51. Aeroelastic solutions generally involve increasing of the structure stiffness or mass balance (passive solutions), which typically involve increase of weight and cost while decreasing performance [4].

17 CHAPTER 1. INTRODUCTION 6 The concept of active control to improve the aeroelastic performance of wings emerged in the fifties [6]. Probably, one of the primary efforts in the direction of active control was the US Active Aeroelastic Wing (AAW) program [7]. The AAW concept is a technology that integrates air vehicle aerodynamics, active controls, and structures together to maximize air vehicle performance. They have played with the wing aeroelastic flexibility by using multiple leading and trailing edge control surfaces, activated by a digital flight control system. The energy of the air was used to achieve the desirable wing twist with very little control surface motion. The AAW concept was successfully tested in the NASA Langley transonic dynamics wind tunnel. Based on these tests, a joint Air Force, NASA and Boeing flight test program was launched [8]. In this program, an F/A-18 fighter was modified to demonstrate the AAW concept. At the end of January 2003, the AAW aircraft had successfully flown eleven research missions [9]. The Russian Aerospace Research Institute tested active aeroelastic concepts using a small additional control surface ahead of the wing leading edge, improving the roll control. They also have developed new structural elements that enable large structural deformations of aerodynamic surfaces, in order to obtain control surface deflections with smooth curvature, thus improving the aerodynamic effectiveness [lo]. In the last two decades, a new actuation concept for structural control has emerged. This concept uses the multifunctional materials properties to control the structural stiffness and shape of the composite materials. Several studies are being performed to demonstrate applications of adaptive structures in aircraft, helicopters and submarines. The adaptive structures technology is expected to significantly reduce dynamic instabilities and vibrations [I 1, 12, 131. In 1990, at the Massachusetts Institute of Technology (MIT), [14], investigations were performed using embedded piezoelectric actuators in laminated materials. In

18 CHAPTER 1. INTRODUCTION 7 Japan, [15], projects have focused in the design of adaptive truss structures. In Europe, researches were performed using shape memory alloys at the University of Twente, Netherlands, and using piezoceramics at ONERA, France. The European Space Agency (ESA) has been investigating the application of smart materials in aerospace structures [16]. In 1991, the the Smart Structures Research Institute was created, at the University of Strathclyde, in Scotland [ly]. In 1998, Forster and Yang [18] examined the use of piezoelectric actuators to control supersonic flutter of wing boxes. The wing box contained piezoelectric actuators that control the twist of the wing, in order to change the free-vibration frequencies and modes, thus, controlling flutter speed. This study has shown that the weight of the wing box can be decreased by adding piezoelectric actuators to meet the flutter requirement at smaller thickness of skins, webs and ribs. In 2003, several studies were developed using the smart structure concepts. For example, the Italian Aerospatial Research Centre (CIRA) designed torsion tubes to produce geometry variations and transmit deformations to mechanic devices. This tube is a cylindrical anisotropic laminated shell. The numerical and experimental results aimed to maximize the tangential rotations and the transmitted energy, in order to obtain suitable deflections of the control surfaces. The main benefits that they observed include the reduction of negative aeroelastic impacts on aircraft performance and stability; cost reduction, by decreasing the size of stabilizer surfaces and total structure weight; reduction of the emissions, by reducing the engine power demand [19]. At the University of Manchester, United Kingdom, in 2003, a research program investigated the development of "active internal structures" concepts, in order to enable the active aeroelastic control of aerospace structures. Using wing internal structures, in particular through changes in the position and stiffness of wing spars, they aimed to control the wing bending and torsional stiffness. Their analytical and

19 CHAPTER 1. INTRODUCTION 8 experimental results showed that it is possible to control the wing twist and bending using this type of internal structures [20]. Also in 2003, at the University of Michigan, USA, a research program has worked to reduce the vibration in a rotorcraft using actively controlled flaps [21]. The active structures concept has also been used in Micro Air Vehicles (MAVs). For instance, at the University of Florida, USA, an investigation has studied the use of morphing as a control effector for a class of MAVs with membrane wings, in the year 2003 [22]. The morphing was restricted to twisting the wing for roll control. Experimental data showed that the morphing can be easily achieved and greatly improves the flight characteristics, when compared with traditional control surfaces State of the Art in Multifunctional Materials Although significant advances in smart materials have taken place in the past decade, the presence of the piezoelectric effect in quartz was experimentally confirmed, over 100 years ago, by Jacques and Pierre Curie [23]. Then, the first application of the piezoelectric crystal effect was force and charge measurement apparatus, patented by the Curies in 1887 [24]. As explained in Section 1.1, piezoelectric materials can be used as sensors and actuators. In sensor mode (called direct mode), the piezoelectric material becomes electrically charged when a mechanical deformation occurs. Piezoelectric sensors can be used in order to detect strain, motion, force, tension and vibrations, since they generate an electric response to these stimuli. In actuator mode (called inverse mode), the piezoelectric material deforms itself when subjected to an electrical field. Piezoelectric actuators can generate motion, force, tension and vibrations. The figure 1.1 illustrates the piezoelectric actuator mode.

20 CHAPTER 1. INTRODUCTION Figure 1.1: Illustration of a piezoelectric actuator deformation. The force and deflection output of the piezoelectric actuators for a given applied voltage can be considered linear, as shown in the figure 1.2. For a given voltage applied to the actuator, its displacement is reduced as the load increases, until the blocking force is reached at zero deflection. On the other hand, the displacement is increased as the load is removed, until the free deflection. The area under the line represents the work done by the piezoelectric actuator. The energy transferred from the actuator to the mechanical system is maximized when the stiffness of the actuator and the mechanical system are matched [25]. It is desirable that piezoelectric sensors have a response that varies linearly with changes in the measured quantity. As a result, piezoelectric elements used in sensors generally operate in the linear region, such that the voltage generated across the element varies linearly with the magnitude of the mechanical stress. For a given piezoelectric material, the amount of voltage produced by the ceramic subjected to a stress can be increased by increasing the thickness of the ceramic [26].

21 CHAPTER 1. INTRODUCTION Figure 1.2: Force-deflection output of a typical piezoelectric actuator. Piezoelectric materials usually have the form of patches, thin disks, tubes or very complex shapes fabricated using solid free form fabrication or injection molding 127, 281. Traditional piezoelectric materials are called PZT (lead zirconate titanate), which have small strain levels (on the order of 0.1% to 0.2% ). The new relaxor ferroelectric single crystals (PZN-PT and PMN-PT) can develop strains on the order of 1% and have approximately 5 times as much strain energy density as conventional piezoceramics [29]. The amount of strain produced in the material is dependent on the thickness of the element and the magnitude of the voltage applied across the thickness. Piezoelectric materials have been investigated to control vibrations and acoustics in a variety of structures [30, 311. For the majority of the piezoelectric actuators, the focus of the research has been on an effort to amplify the deflection of the material. Piezoelectric actuators can be classified in three different categories, based on its amplification scheme: internally leveraged, externally leveraged, and frequency leveraged. Internally leveraged actuators generate amplified strokes through their internal structure without using external

22 CHAPTER 1. INTRODUCTION 11 mechanical components, including: bender, stack, reduced and internally biased oxide wafers (RAINBOW), composite unimorph ferroelectric driver and sensor (THUN- DER), telescoping, C-block, Recurve and Crescent actuators. Externally leveraged actuators are based on external mechanical components to achieve their actuation ability, including: flexure-hinged, Moonie, Cymbal, double-amplifier, bimorph-based, pyramid, X-frame, and flextensional hydraulic actuators. Frequency leveraged actuators depend on an alternating control signal to generate motion, including inchworm and ultrasonic motors. Among the internally leveraged actuators, stack actuators are thin piezoceramic patches piled in order to linearly increase their overall deflection, and maintaining a low voltage requirement (see figure 1.3(a)). Displacement and force of a stack actuator are directly proportional to its length and cross-sectional area, respectively. Another internal leveraged actuator is the bender. Bender actuators are composed by two or more layers of piezoelectric material, which are poled and activated such that layers on opposite sides of the neutral axis have opposing strains. These opposing strains of the two piezoelectric layers create a bending moment, causing the entire bender to bend. In order to achieve structural stability, inactive substrates may be added to these active layers. As an illustration, most benders have piezoelectric material extending the full length of the beam, as shown in figure 1.3(b). Some practical difficulties related with the use of raw piezoceramics as actuators include soldering, cracking because of their fragile nature, and electrical isolation. The QuickPack actuators (manufactured by ACX Inc.) are a step forward in terms of applying piezoelectric technology to commercial products (see figure 1.4). This is the type of actuators used to perform the experiments focused on this thesis. These actuators contain two piezoceramic elements enclosed in a protective polyrnide insulation material. The QuickPack actuators can be used as patches, to induce in-

23 CHAPTER 1. INTRODUCTION Figure 1.3: (a) Diagram of a piezoceramic stack; (b) Typical bimorph bender actuator. plane strain, or operated out of phase to act like a bimorph bending actuator (similar to the bender actuator). Quickpack actuators eliminate the need of soldering leads to the piezoelectric material, improve the durability of the actuator, and electrically isolate the actuator from the attached surface. A unimorph bender is a special case of a bender actuator. It is a composite beam, plate or disk with one active layer and one substrate (an inactive layer). The mentioned RAINBOW, Crescent and THUNDER actuators are typically referred to as unimorph benders. The RAINBOW actuator is a piezoelectric wafer that is chemically reduced on one side. A partially metallic layer is formed on one side placing the piezoelectric element in compression, forming a hemispherical container. The Crescent actuator is very similar to the RAINBOW actuator. It is a stressedbiased unimorph actuator that is fabricated by cementing (using epoxy or solder) metal and electroded ceramic plates together, at an elevated temperature. When the actuator approaches the room temperature, a prestress is induced in the active material (due to the difference in the coefficients of thermal expansion of the metal

24 CHAPTER 1. INTRODUCTION Figure 1.4: ACX Quickpack Actuator. and the ceramic) resulting in a unimorph actuator that is curved in shape. The THUNDER is another prestressed actuator that consists of a layer of a ceramic wafer attached to a metal backing using a polymide adhesive film. The C-block, Recurve and telescoping actuators are called building-block actuators. Building-block actuators have numerous small actuation units, called building blocks, which are combined in series and/or parallel to form larger actuation systems with improved performance. Externally leveraged actuators use an external mechanism to increase the output deflection, by decreasing the output force. These external mechanisms can be mechanical or hydraulic. A simple way of increasing the displacement of an actuator is the use of a mechanical lever arm. Although this mechanism increases the displacements output, the actuator force is decreased. The flexure-hinged actuator uses this principle. Flextensional actuators use a piezoceramic stack and an external amplification mechanism, in order to convert the motion generated by the stack to a usable output motion in the transverse direction. Moonie, Cymbal and bimorph-based are examples of flextensional actuators.

25 CHAPTER 1. INTRODUCTION 14 In the category of the frequency leveraged actuators, the output strain of the actuators is increased by using the frequency performance of the piezoelectric material to rapidly move the actuator in one direction in a series of small steps. The first type of actuator to operate using frequency was the Inchworm. The Inchworm consists of three connected actuators that actuate in sequence to move the actuator down a rod. Besides piezoelectric materials, there are more multifunctional materials that can be used in order to obtain an adaptive structure. Next, it will be presented some of these multifunctional materials. Shape memory alloys (SMA) return to their original form when heated above their critical temperature, i.e., they "remember" their original crystalline structure or shape. SMAs are used as actuators to change characteristics of the host structure. They have relatively large actuation force and high strain output damping capabilities. However, they may have large hysteresis, and due to their slow response, they are best suited for low frequency applications. The most common shape memory alloy is nitinol. Some efforts were made in order to use shape memory alloys as actuators, including embedded shape memory alloys in composites or in conventional structures. Some studies were performed using SMA, like the development a model that studies the behaviour of composites with shape memory alloys [32]. SMAs are also quite suitable for slow motion of control surfaces such as flaps in helicopters [33]. Electrostictive materials are similar to piezoelectric materials. The electrostrictive actuators are characterized by possess the electrostiction property. Electrostiction is a phenomenon observed in all dielectric materials. When an electric field is applied across the dielectric material, the dipoles align themselves with the field. This process induces an internal strain and the material changes its dimensions [34]. When the electric field is removed the dipoles re-orient and the material returns to its original dimensions. The induced strain is proportional to the square of the applied electric

26 CHAPTER 1. INTRODUCTION 15 field, thus it is always positive, i.e., the material is under tension. The most popular electrostictive material is the lead magnesium niobate (PMN), which have high strain capabilities and very low hysteresis properties. The magnetostrictive materials modify their dimensions when a magnetic field is applied. Their dimension change is a result of a re-orientation of the atomic magnetic moments, or small magnetic domains. As the magnitude of the applied magnetic field increases, more domains become aligned, until magnetic saturation occurs (when all magnetic domains are aligned with the applied magnetic field). When the magnetic field is removed, the material returns to its original dimensions. The produced strain is proportional to the square of the magnetic field. Thus, like the electrostrictive materials, the strain is always positive (tension). Terfenol-D is one of the most popular magnetostrictive materials. Studies [35] have demonstrated that strains produced by magnetostrictive materials are smaller than those produced by electrostrictive materials, and the hysteresis is higher than the electrostictive material. The rheological fluids are multiphase materials that consist of field-responsive particles suspended in a carrier non-conducting fluid [36]. They can be electro- or magneto-rheological fluids (ER or MR). The viscosity of ER and MR fluids varies when an electric or magnetic field, respectively, is applied. These active fluids can adapt and respond almost instantly and have been used in clutch, brake, valve-type devices, dampers, and shock absorbers [37, 381. Optical Fibre sensors respond to strain and temperature when a shift in their optical wavelength takes place. They can be used in the structures skin or directly embedded in the structure. Many optical fibres can be manufactured onto a single optical fibre, and then interrogated independently to provide distributed measurements over large structures, such as civil infrastructures and ships [3].

27 CHAPTER 1. INTRODUCTION Structure of the Thesis Prior presenting the attained results in the different performed analyses, some fundamentals of flutter are presented in first Section of the Chapter 2 of this thesis. This Section describes the fundamentals of structural vibration, in which is presented the governing equation of a general vibrating system, and its basic elements. The general methods for the determination of the natural frequencies and mode shapes of a system are identified, respectively, as eigenvalue and eigenvector extraction problems. The relation between the wing vibrations and wing flutter speed is explained. The flutter phenomenon is defined and some different types of flutter are presented. The equation of motion of an aeroelastic system, in terms of a discrete system, is explained. Finally, a brief explanation of the ZAERO flutter solution technique is presented. In Chapter 2, the remaining Sections have the objective to present the results of the computational analyses performed in this project. Two different computational analyses were performed: a finite element analysis, using ANSYS program, and a flutter analysis, using ZAERO program. The finite element analysis is presented in Section 2.2, and was done in order to generate the passive wings natural frequencies and mode shapes. Note that only the passive wings were studied. No ANSYS analysis was performed using the wings in active configuration. In Section 2.3 describes the flutter analysis. The ZAERO program was used, which imports the solution generated by ANSYS program and calculates the wings flutter modes and speeds of the wings in passive configuration. The ZAERO program was used to calculate the flutter speed only when the wing has a no conventional configuration. In the case of a conventional wing, i.e. with a main spar and ribs, a method presented by [39] is used to calculate the flutter speed.

28 CHAPTER 1. INTRODUCTION 17 The wind tunnel tests are presented in Chapter 3. First, all the experimental apparatus is described in Section 3.1. In this Section the following hardware components are presented: the wind tunnel, the tests articles, the digital controller, and electronic equipment. In the tests articles Subsection, an extensive description of the respective articles is performed. Several photographs of the articles are shown, and several wing schemes were included in order to explain the active wings control vibration process. In Section 3.2 the wind tunnel tests objectives and procedures are described. In this Section, the entire hardware system used for wind tunnel tests is presented, explaining all the connections in the circuit. Additionally, the approach followed in order to design the control model, and the control model itself, are presented. Yet in Section 3.2, it is performed the description of the tests procedures and measured data. Finally, in Sections 3.3 and 3.4 the wind tunnel results of, respectively, the active skin wing and active spar wing are presented. The Chapter 4 describes the performed flight tests. The hardware involved in these tests is presented in Section 4.1. This Section includes the description of the RPV flight model and the additional electronic equipment needed exclusively for the flight tests. Like in Section 3.2, in Section 4.2 the flight tests objectives and procedures are described. The control approach used for these tests is also explained. At the end of this Chapter, in Section 4.3, the flight tests results of the RPV with active spar wing are presented.

29 Chapter 2 Computational Analyses 2.1 Fundamentals of Flutter As referred in Chapter 1, any motion that repeats itself after an interval of time is called vibration or oscillation [40]. A system that vibrates is generally defined through three properties: elasticity, a mean of storing potential energy; mass or inertia, a mean of storing kinetic energy; and damping, a mean of losing energy. In vibration theory, a vibrating system is generally described with the following elements: mass, spring or stiffness, damper and excitation force. This way, the first three elements describe the physical vibrating system. The mass and spring store energy and the damper dissipates it in the form of heat. On the other and, the energy is given to system by the excitation force. Thus, a single degree of freedom system, with translational mass, is generally defined as shown in figure 2.1. In this figure, k is the spring constant, c is the damper constant, m is the mass, f (t) is the excitation force, and x(t) is the mass displacement.

30 CHAPTER 2. COMPUTATIONAL ANALYSES Figure 2.1: Single degree of freedom system with translational mass. In terms of the type of excitation force, the vibration is known as free vibration when the system, after an initial disturbance, is left to vibrate on its own. On the other hand, the vibration is called forced vibration when the system is often subjected to an external force. In terms of damping, an undamped vibration happens if no energy is dissipated in friction or other resistance during the vibration; a damped vibration occurs if there is energy lost during oscillation. In terms of periodicity, a system has a deterministic vibration if the magnitude of the excitation acting on the system is known, or random vibration if that magnitude can not be predicted. The vibration of a wing is a random vibration since it is excited with the wind. A degree of freedom of a system is the minimum number of independent coordinates required to determine the positions of all parts of this system, at any time. Systems with a finite number of degrees of freedom are called discrete systems. Systems with an infinite number of degrees of freedom are usually called continuous systems. Most of the time, continuous systems are treated as discrete systems, since the methods to analyze continuous systems are only applicable to simple problems. In this study, our system is a wing, and because all its components has an infinite number

31 CHAPTER 2. COMPUTATIONAL ANALYSES of mass points, thus has an infinite number of degrees of freedom. Considering the wing system, the governing equation may be represented as follows: The case presented in figure 2.1 is a very simple vibrating case, in which it is possible to analytically determine the exact solution. However, the solution of the systems governing equations is often more complex in real problems, and it is impossible to consider all the details for the mathematical model, or because they have infinite number of degrees of freedom or they have considerable irregularities in the oscillatory motion. A normal procedure is the use of numerical methods involving computers to solve these equations. For instance, the finite element method is a numerical method that can be used for the accurate solution of complex structural vibration problems. This is a numerical method in which the structure is divided and replaced by several pieces or elements. Each one of these elements is assumed to be a continuous structural member called finite element. All the elements defining the structure are assumed to be interconnected at certain points known as nodes. During this method solution process, the equilibrium of nodal forces and the compatibility of displacements between the elements are satisfied, such a way that the entire structure is made to behave as a single entity [40]. Because of the complexity of the structures in study, it was previously decided that a commercial code would be used in order to calculate the solution of the problem. In this thesis, the wings structures were studied using the ANSYS program, which has many finite element analyses capabilities. Having presented the basic elements of a vibrating system and its general governing equation (2.1), it is now time to describe the type of vibration analysis that was done in this thesis. Since the final objective is to calculate the wings flutter speeds, the vibration analysis that was performed consists in the determination of

32 CHAPTER 2. COMPUTATIONAL ANALYSES 21 the systems natural frequencies and mode shapes, which in ANSYS program is called modal analysis. Since the determination of systems natural frequencies is an eigenvalue problem, which solution corresponds to the undamped free vibration of the system, the ANSYS modal analysis starts by solving the following equation of motion: Then, for a linear system, free vibrations will be harmonic of the form: where (4)i is the eigenvector representing the mode shape of the ith natural frequency, wi is the ith natural circular frequency, and t = is the time. Thus, equation (2.2) becomes: (-wi2[m1 + [kl){4)i = (0) (2.4) This equality is satisfied if either {$)i = (0) or if the determinant of ([k]- wi2[m]) is zero. The first option is the trivial one and, therefore, is not of interest. Thus, the second one gives the solution: This is an eigenvalue problem, which may be solved for up to n values of w2 and n eigenvectors {q5)i, which satisfy the equation (2.4), where n is the number of degrees of freedom of the system. The ANSYS has several techniques to perform the eigenvalue and eigenvector extraction. In this project the Subspace Method was used to perform this extraction, and its algorithm description is available in the ANSYS manual. There is a strong relation between the vibration of a wing and its flutter speed.

33 CHAPTER 2. COMPUTATIONAL ANALYSES 22 It is known that a wing in flight is in continuous vibration because of the air flow. Additionally, when the wing is disturbed by the wind (for instance, when a gust strikes the wing), the wing motion may be such that the amplitude of vibration tend to decrease, remain constant or increase. The first case occurs when the airspeed is between zero and the critical flutter speed (stable condition). The wing vibration will remain constant when the airplane is flying at flutter speed (neutral stability). Finally, at speeds higher than the flutter speed, the wing vibration tends to increase, i.e., divergent oscillations take place, which may cause destruction of the wing. It should be stated that the aerodynamic forces which tend to maintain the wing vibrations exists because of the wing vibrations themselves. Thus, the flutter can be defined as an aeroelastic, self-excited vibration, in which the external source of energy is the air stream [41]. The classical type of flutter is called classical flutter and involves the coupling of several degrees of freedom of the structure. A typical example of wing classical flutter is the called wing bending-torsion flutter, which has coupling between bending and torsion mode shapes. It is known that oscillations caused by pure bending or pure torsion modes are rapidly damped. However, when there is a coupling between bending and torsion oscillations, the aerodynamic and inertial forces acquire an unstable effect. The non-classical type of flutter involves only one degree of freedom of the structure, and the stall flutter and aileron buzz are some examples. Also, it is important to remember that the flutter phenomenon, as with all aeroelastic phenomena, is highly sensitive to the structures vibration modes, depending on their natural frequencies and mode shapes. In this thesis, again because of the complexity of the structures in study, it was decided that a commercial code will be used to calculate the flutter speed and characteristics. The program used to perform these tasks was the ZAERO, which integrates the essential disciplines required by aeroelastic design and analysis. This program im-

34 CHAPTER 2. COMPUTATIONAL ANALYSES 23 ports the solution of the free vibration previously generated by the ANSYS solution. Essentially, when this importation is made, ZAERO is importing relevant information about the structural mesh of the structure, the natural frequencies and mode shapes, the mass and stiffness matrices generated by the structural finite element method. The ZAERO program incorporates two different techniques to determine the flutter solution: the K-method and the g-method. The K-method is performed at a given Mach number, M, and presented in terms of velocity versus frequency diagram ( V-f diagram) and velocity versus damping diagram ( V-g diagram). This method requires only a straightforward complex eigenvalue analysis of each reduced frequency, thus its solution technique is efficient and robust. However, the frequencies and velocities are computed at a given pair of Mach number and air density. This implies that the flutter boundary computed by the K-method generally is not a "matched point" solution in that the flutter velocity, Vf # Ma,. The matched point solution can be achieved only by performing the flutter analysis at various air densities iteratively until the condition of Vf = Ma, is satisfied. For n structural modes, the K-method normally provides only n roots of the flutter equation. However, the number of roots could exceed the number of the structural modes. Unlike the K-method, the g-method potentially gives an unlimited number of roots, which could provide important physical insight of the flutter solution. More information about this methods and respective mathematical algorithms can be seen in [42]. Next, the fundamentals of aeroelasticity that ZAERO uses are briefly presented. The aeroelastic response of an aircraft in flight is the result of the interaction of inertial and elastic structural forces, aerodynamic forces induced by the structural deformations and external disturbance forces. Thus, the equation of motion of an aeroelastic system, in terms of a discrete system, can be derived based on the equi-

35 CHAPTER 2. COMPUTATIONAL ANALYSES librium among these forces, as follows: where [m] and [k] are, respectively, the mass and stiffness matrices generated by the structural finite element method, performed by ANSYS, {x(t)) is the structural deformation, and { f (t)) represents the aerodynamic forces applied on the structure. In general, {f(t)) can be divided in two parts: external forces, {fe(t)), and aerodynamic forces induced by the structural deformation, {fa(t)), i.e.: The ZAERO program considers that external forces acting on the system are provided by the user. Typical examples of external forces are atmospheric turbulence and impulsive type gusts. The generation of the aerodynamic forces is based on the theoretical prediction that requires the unsteady aerodynamic computations, and depends on the structural deformation {x(t)). Thus, ZAERO program solves the following equation: that can be represented by the diagram in figure 2.2. The left hand side of the equation (2.8) is a closed-loop aeroelastic system which can be self-excited in nature. This gives rise to a stability problem of the closed-loop aeroelastic system known as flutter. If {fa (t)) is a nonlinear function with respect to {x(t)), the flutter analysis must be performed by a time-marching procedure solving

36 CHAPTER 2. COMPUTATIONAL ANALYSES Figure 2.2: Aeroelastic functional diagram. the following equation: However this time-marching procedure is computationally heavy, since it requires a nonlinear time-domain unsteady aerodynamic method. The ZAERO practice of flutter analysis consists in remodel equation (2.9) into a set of linear systems and to determine the flutter boundary by solving the complex eigenvalues of the linear systems. This procedure is based on the assumption of amplitude linearization, which states that the aerodynamic response varies linearly with respect to the amplitude of the structural deformation. This way, the flutter analysis becomes a eigenvalue problem. In this case, the aerodynamic system can be approximated by a linear system for which an aerodynamic transfer function (that relates the aerodynamic feedback fa(t) with the structural deformation x(t)) can be defined. Knowing this transfer function, equation (2.9) can be transformed into the Laplace domain and results in an eigenvalue problem in terms of s, i.e.:

37 CHAPTER 2. COMPUTATIONAL ANALYSES 26 where q,h represents the aerodynamic transfer function, q, is the dynamic pressure, L is the reference length (is generally defined as half of the reference chord), and V is the velocity of the undisturbed flow. More details about the ZAERO approach and methods are available in [42]. 2.2 Finite Element Analysis As referred in the previous Section, the analysis of free vibration using the finite element method was performed in order to generate the passive wing natural frequencies and mode shapes. This study was performed using ANSYS program. The intent of the finite element solutions and flutter calculations, presented in the following Sections, is only to determine the wing natural frequencies, mode shapes, and flutter speeds in passive configuration, i.e., without the actuation of the piezoelectrics. It is only in Chapter 3 that both the passive and active wings solutions will be studied and compared with each other. In this analysis, complete three-dimensional wings analyses were performed, i.e., all the wings components were defined using two-dimensional elements in which the thickness is given. All the wing components (see wings description in Section 3.1.2), like the two carbon fibre plates, ribs, leading and trailing edges, were defined using the SHELL93 8-node structural shell elements. This ANSYS finite element type is proper to model curved skinned components of orthotropic materials, which is the present case (see materials properties in the table 3.1).

38 CHAPTER 2. COMPUTATIONAL ANALYSES Adaptive Skin Wing The designed final wing finite element model had nodes, i.e., degrees of freedom, and in figure 2.3 it can be seen the wing final mesh. Figure 2.3: Adaptive Skin wing finite element mesh, defined in the ANSYS program. In terms of constrained nodes, it was considered that the nodes defining the holes of the two balsa wood ribs, one placed on the wing root and another at the distance of 20 cm relatively to the wing root, had zero linear displacements and rotations. Note that, in the real wing, those holes define the connection between the wing and the fuselage (see Section 3.1.2). After running the ANSYS Modal Analysis task, the first ten natural frequencies and mode shapes were extracted, and the results are displayed in table 2.1 and figure 2.4, respectively. Analyzing these results, it was obvious that this wing does not have a conventional behavior in terms of mode shapes, since several shell local vibrations were found. Note that this result should be expected, since this wing does not have a conventional configuration. Conventional wings have a spar as the main stiff component, and not a st8 skin. Because of this non-conventional wing configuration, a

39 CHAPTER 2. COMPUTATIONAL ANALYSES 28 non-classical type of flutter should be expected for this wing. The mode shapes that lead with shell vibrations are: the 2nd, 4th, and 8th modes, with shell bending, and the 7th mode, with shell bending-torsion. This way, the 2nd, 4th, 7th and 8th modes are mainly characterized by the two carbon plates vibration in which the lower and upper carbon plates have deformations in opposite phases. This can be explained due to the fact that the two carbon plates are only connected with each other at some discrete points, along the leading and trailing edges. These modes are not beneficial to this study, since the main objective is to control the vibration of the total wing as a single assembly and not the lower or the upper carbon plates separately. See also Section in order to read additional information about this point. In Section 3.3.1, the wind tunnel tests proved that the problem which caused these local shell vibrations was related with a few number of connections between the lower and upper carbon plates. On the other hand, the remaining modes have conventional behavior, as follows: lst mode is the first bending, 3'd mode is the first bending-torsion mode, 5th mode is the first torsion mode, 6th mode is the second bending-torsion mode, gth mode is the second torsion mode, and loth mode is the third bending-torsion mode. Table 2.1: Adaptive Skin wing first ten natural frequencies. Mode Natural Frequency [Hz]

40 CHAPTER 2. COMPUTATIONAL ANALYSES Figure 2.4: Adaptive Skin wing first ten mode shapes.

41 CHAPTER 2. COMPUTATIONAL ANALYSES Adaptive Spar Wing As in the adaptive skin wing, all the adaptive spar wing components (see Section 3.1.2), i.e., the carbon fibre beam, ribs, leading and trailing edges, were defined using the SHELL93 &node structural shell elements. The final wing finite element model had 4822 nodes. In figure 2.5 it can be seen the wing final mesh. Figure 2.5: Adaptive Spar wing finite element mesh, defined in the ANSYS program. Using the ANSYS Modal Analysis task, the first five natural frequencies and mode shapes were extracted, and the obtained results are displayed in table 2.2 and figure 2.6, respectively. Since this passive wing has a conventional configuration, Table 2.2: Adaptive Spar wing first five natural frequencies. Mode Natural Frequency [Hz]

42 CHAPTER 2. COMPUTATIONAL ANALYSES Figure 2.6: Adaptive Spar wing first five mode shapes.

43 CHAPTER 2. COMPUTATIONAL ANALYSES 32 i-e., with a main beam, internal ribs and a non-stiff skin, conventional flutter was expected. Analyzing the results, no shell local vibrations were found, and the following natural modes were determined: lst mode is the first bending, 2nd mode is a "swing" mode (bending in the wing plane), 3rd mode is the first torsion, 4th mode is a bending-torsion mode, and 5th mode is the second torsion mode. Since this wing has a conventional configuration, classical flutter (referred in Section 2.1) is expected, i.e., wing bending-torsion flutter. 2.3 Flutter Analysis Adaptive Skin Wing After obtaining the ANSYS results, the ZAERO program was used to perform the wing aeroelastic study in terms of flutter. ZAERO program imports the solution of the free vibration generated by the ANSYS program. The flutter analysis was performed using a constant air density of 1.225Kg/m3, and flutter speeds between 15 and 250mls were calculated. Table 2.3 shows the flutter results obtained using the g-method. The first three columns on the left, reading from the left to the right, show the first six flutter modes, their respective speeds and frequencies. The column on the right displays the natural modes contributions for the flutter ocurrence, in each flutter mode. The results in table 2.3 state that: - the lst flutter mode, at 44.87m/s, is essentially related with the wing 7th natural mode; - the 2nd flutter mode, at 60.44m/s, happens because of the coupling between the wing 3'd and 2nd natural modes;

44 CHAPTER 2. COMPUTATIONAL ANALYSES 33 - the 3'd flutter mode, at m/s, is related with the coupling of the wing loth and gth modes; - the 4th flutter mode, at m/s1 occurs because of the strong coupling among the gth, gth and loth natural modes; - the 5th flutter mode, at m/s, is dependent of the lst, 7th, 3rd, 2nd and gth wing natural modes; - the 6th flutter mode, at m/s, occurs when the coupling among 7th, 6th and 3'd natural modes takes place. Table 2.3: Adaptive Skin wing flutter results, using the g-method. Flutter Mode Speed [m/s] Frequency [Hz] Natural Modes contribution [%] 1)0.28, 2)1.49, 3)2.81, 4)0.12, )5.48, 6)15.53, 7)100.00,

45 CHAPTER 2. COMPUTATIONAL ANALYSES 34 Additionally, the V-f (velocity versus frequency) and V-g (velocity versus damp- ing) graphics were obtained, as shown in figures 2.7 and 2.8. Finally, the figure 2.9 shows the obtained six flutter modes shapes Speed [ds] Figure 2.7: Adaptive Skin wing V-f graphic. Analyzing the V-g graphics in figure 2.8, one can conclude the following: the wing natural mode that become unstable first is the 7th mode, at 44.87mls speed; then is the 3'd mode, at 60.44mls speed; after that, it follows the gth, loth, 4th and 6th modes at, respectively, m/s m/s mls and mls speeds. Note that these found flutter modes do not have all the same intensity. In fact, the first flutter mode to appear, related with the wing 7th natural mode, as a smoother behaviour and smaller unstable damping values (i.e., positive damping values smaller than 0.025) when compared with other flutter modes. For instance, the second flutter mode, related with the wing 3Td natural mode, has an abrupt behaviour since high unstable damping values (i.e., positive damping values higher than 0.025) are suddenly reached.

46 CHAPTER 2. COMPUTATIONAL ANALYSES D " a hr2ade;l -0.6 *I~c~EI - m m =-wm 5 Speed [mls] *Mule6 +Mode7 - Mc&8 +MEds!9 +Mde Speed [mis] Figure 2.8: (a) Adaptive Skin wing V-g graphic; (b) V-g graphic zoom near the zero damping.

47 CHAPTER 2. COMPUTATIONAL ANALYSES Flutter mode 1 Flutter made 2 Flutter mode 3 Figure 2.9: Adaptive Skin wing flutter modes shapes.

48 CHAPTER 2. COMPUTATIONAL ANALYSES 37 This means that the flutter mode related with the 3rd natural mode is a more abrupt and dangerous phenomenon. Of course, one should not forget that the flutter mode related with the 7th natural mode is the one which first occurs. In figure 2.9 it is possible to confirm that all the obtained flutter modes are related with shell vibrations. The first and third flutter modes shapes have, essentially, shell vibration. The second flutter mode shape also have bending, the fourth have torsion, and the fifth and sixth flutter modes have bending-torsion Adaptive Spar Wing As referred in Section 1.3, the flutter speed of wings with conventional configuration, i.e. with a main spar and ribs, was calculated using a method presented by [39]. This method is based on an empirical investigation and calculates the torsional flutter of gliders and small aircrafts wings. First, it is necessary to estimate the wing first torsion frequency using vibration tests. However, the wing first torsion frequency was already calculated in the ANSYS analysis presented in Section 2.2.2, and its value is ft = 46.66Hx. Thus, applying the equation 2.11 it is possible to estimate the wing flutter speed. Later, in Chapter 3, the flutter speed of this wing will be calculated using the wing experimental tests results. where VT is the flutter speed in m/s, ~0.7 is the wing chord length at 0.7b/2 in m, in which b is the wing span, ft is the wing first torsion frequency in Hz, and A is the wing aspect ratio. Since the wing is rectangular, i.e. has a constant chord, thus ~ 0.7 = c = 0.33m. The wing aspect ratio is A = (see table 4.1). The equation 2.11 gives an

49 CHAPTER 2. COMPUTATIONAL ANALYSES 38 empirical estimation of flutter speed for wings with aspect ratios between 6 and 9, which is the present case. Using the referred values of ~0.7, ft and A in equation 2.11, the wing estimated flutter speed is VT = 49.83rnls.

50 Chapter 3 Wind Tunnel Tests 3.1 Experimental Apparatus The hardware involved in these wind tunnel tests is described in four Sections: the wind tunnel, the tests articles, the digital controller, and electronic equipment Wind Tunnel The RPV model with the adaptive wings was tested in the wind tunnel installed at Portuguese Air Force Academy Aeronautical Laboratory, shown in figure 3.1. The wind tunnel is a closed circuit horizontal wind tunnel with a maximum operating velocity of 70m/s, with the air stream temperature control. The test section is 1.3m x 0.8m and 2m long, and can be used in an open or closed configuration. A uniform flow velocity with less than 0.8% in pressure variation can be obtained in a cubic zone 1. lm x 0.6m x 1.4m. The ventilator has a 8 blades fan with 1.6rn diameter. The shaft has a 0.63m diameter and rotates at maximum speed of 1600rprn, producing a maximum airflow of 72.8rn3/s, with 88% efficiency.

51 CHAPTER 3. WIND TUNNEL TESTS Figure 3.1: Photograph of the Air Force Academy Aeronautical Laboratory wind tunnel. The RPV has a total span of 2.40m and the wind tunnel tests were performed with half of the RPV model (1.20m span), in open section configuration. The figure 3.2 shows the test article mounted in the wind tunnel section Tests Articles The RPV with active wings wind tunnel model consists of four main components: the RPV fuselage to mount the half wing, the half wings, the piezoelectric sensors and the piezoelectric actuators. There were three driving factors in the design of the RPV model: half RPV model had to fit inside the wind tunnel section, the wing had to flutter within the wind tunnel envelope and had to have surfaces (in the skin for the active skin wing, and in the main spar for the active spar wing) on which piezoelectric sensors and actuators could be mounted. The RPV fuselage, shown in figure 3.3, has a 0.4m span, 1.4m in length, and 0.2m in width. It has two fixation points to perform the connection between the fuselage and the half wing. The RPV fuselage is settled in a vertical stationary panel that can

52 CHAPTER 3. WIND TUNNEL TESTS Figure 3.2: Photograph of the RPV model with active wing tunnel section. mounted in the wind rotate in order to allow the variation of the RPV angle of attack. Both wings (active skin and active spar) have 2m of total length, which added with the 0.4m of fuselage span gives the total RPV span of 2.4m, and 0.33m chord. The RPV fuselage and half wing are joined together by means of a main rod tube and an auxiliary leading rod, as shown in figure 3.4. The active skin wing has a modified NACA0012 airfoil. The wing leading edge skin is made of fibre glass 0.75mm thick, with eight distributed balsa wood leading edge ribs 2mm thick along the wing span, to assure that the leading edge shape is maintained. The wing trailing edge is balsa reinforced and corresponds to the wing flaperon. The active structural elements of the wing are two flat sheets of unidirectional &layered carbon fibre, with a total thickness of 1.104mm (the upper and lower surfaces of the wing), used as the skin of the wing. The carbon layers are oriented along the span direction. The leading and trailing edges are both attached to the carbon fibre flat sheets using fibre glass brackets, as can be seen in figure 3.5. The whole wing has only three integral ribs: one in the wing root, one at the distance of 20cm relatively to the wing root (important to make the connection with

53 CHAPTER 3. WIND TUNNEL TESTS Figure 3.3: Photograph of the RPV wind tunnel model fuselage. Figure 3.4: Schematic view of the interconnection between the RPV fuselage and wing (not to scale).

54 CHAPTER 3. WIND TUNNEL TESTS Table 3.1: Wing material properties. Property [Unity] Carbon Fibre Balsa Wood Fibre Glass Ex [GPa] EY [GPal EZ [GPa] Gz, [GPa] Gy, [GPa] Gx, [GPa] "XY "YZ Zwm3~ Carboa fibre flat plates Leading edge Fibre glass hrackeis Trailing edge Figure 3.5: Schematic view of the wing airfoil shape and components. the fuselage), and another in the wing tip. These ribs are made of balsa wood 3mm thick. The materials properties of the active skin wing components are shown in table 3.1. In the wing lower surface the fibre glass brackets are glued to the carbon flat sheet. In the wing upper surface, those components are connected by bolts, to permit the posterior disassembly of the wing. The figures 3.6 and 3.7 show, respectively, the internal parts of the wing without and with the piezoelectric mounted in the skin. The piezoelectric sensor was mounted near the wing root, in the internal part of the lower carbon fibre plate. The four piezoelectric actuators were mounted in the internal pajrt

55 CHAPTER 3. WIND TUNNEL TESTS Figure 3.6: Photograph of active skin wing internal assemblage (without piezoelectric~). Figure 3.7: Photograph of active skin wing with piezoelectric sensors and actuators mounted in the wing lower surface. of the two carbon fibre plates. The sensor and actuators location scheme is in figure 3.8. The location of the piezoelectric components was decided from the finite element analysis of the vibration modes, which revealed the points with larger strain values. In this wing, is important that the two carbon plates have similar displacement. For instance, when the upper plate is deflecting down, the lower plate should also be deflecting down. This way, when the wing is deflecting down, the upper plate extends and the lower plate contracts. In this case, the two carbon plates work like "active skins" when the piezoelectric actuators placed in the upper plate contract and the piezoclectric actuators placed in the lower plate extend, forcing the wing to deflect

56 CHAPTER 3. WIND TUNNEL TESTS Figure 3.8: Schematic view of the piezoelectric sensor (small patch) and actuators (big patches) placement inside the active skin (gray panel) wing. Figure 3.9: Photograph of the RPV model with active skin wing in the wind tunnel.

57 CHAPTER 3. WIND TUNNEL TESTS 46 in the opposite direction. This process can be seen in figure 3.10(a). On the other hand, when the wing is deflecting up, the upper plate contracts and the lower plate extends. In this case, the two carbon plates work like "active skins" when the upper piezoelectric actuators extend and the lower piezoelectric actuators contract, forcing the wing to deflect down. In figure 3.10(b) it can be seen an schematic view of this process. The figure 3.9 shows the RPV with active skin wing ready to start the wind tunnel tests. Like the active skin wing, the active spar wing has lm of span, giving 2.4m of span to the RPV. The wing has a FX airfoil, which is one of the most desirable airfoils for high-lift low Reynolds models. This wing has a hollow squared beam, which has 1.104mm thickness carbon fibre upper and lower horizontal surfaces (caps), and 3mm thickness balsa wood vertical surfaces (webs), as shown in figure The active elements are the two horizontal carbon fibre plates, which has the same material properties as the carbon fibre used for the active skin wing surfaces. The ribs and the D cell that form the leading edge are made of balsa wood with 2mm thickness. The trailing edge is made of high density balsa wood. The materials properties of the active spar wing components are in the table 3.1. The figures 3.12 and 3.13 show, respectively, the internal parts of the wing without and with the piezoelectric sensors and actuators mounted in the main beam. Like in the active skin wing, the piezoelectric sensor was mounted near the wing root, in the internal part of the lower carbon fibre plate, and the four piezoelectric actuators were mounted in the internal part of the two carbon fibre plates. The sensor and actuators location scheme is shown in figure Also, the location of the piezoelectric components was decided from the previous finite element analysis of the vibration modes. In this wing, since the caps are connected with each other by the webs, the hollow squared beam has an overall desired movement, deflecting up and down. When the

58 CHAPTER 3. WIND TUNNEL TESTS Figure 3.10: Scheme of the active skin wing control vibration process (the rectangular patches represent the piezoelectric actuators and the gray panels the carbon fibre plates).

59 CHAPTER 3. WIND TUNNEL TESTS Figure 3.11: Schematic view of the wing airfoil shape and hollow squared beam. hollow squared beam is deflecting down, the upper cap extends and the lower cap contracts. In this case, the hollow squared beam works like an "active spar" when the piezoelectric actuators placed in the upper cap contract and the piezoelectric actuators placed in the lower cap extend, forcing the spar to deflect in the opposite direction. On the other hand, when the hollow squared beam is deflecting up, the upper cap contracts and the lower cap extends. In this case, the beam works like an "active spar" when the upper piezoelectric actuators extend and the lower piezoelectric actuators contract, forcing the spar to deflect down. The process is similar with the one explained for the active skin wing, but now the spar vibration control is being performed instead of controling the vibration of the skin. Figure 3.2 shows the RPV with active spar wing ready to start the wind tunnel tests. The piezoelectric sensors used for this study were the lead-zirconate-titanate piezoelectric patches, shown in the figure This sensor type measures strain changes, as extensometers do, and are relatively easy to glue to a surface. The piezoelectric sensors are 22 x 38[mm] patches that become electrically charged when subjected to a mechanical strain, producing a variable f 2.5V AC electrical signal. The piezoelectric actuators used were the ACX Quickpack 40W, shown in the figure 1.4, that are rectangular patches of 102 x 40[mm]. This actuator is built through

60 CHAPTER 3. WIND TUNNEL TESTS Figure 3.12: Photograph of active spar wing internal assemblagc (without piezoelectric~). Figure 3.13: Photograph of active spar wing with piezoelectric sensors and actuators. a proprietary manufacturing process that shields the piezoelectric material in a psotective polyamide coating with pre-attached electrical leads and quick connectors, improving electrical isolation and adding protection against breakage during assembly and resistance to micro cracks during operation. The manufacturer guarantees a proper frequency response of the actuator for input signals with frequencies between 1 Hz and 20kHz, and the excitation signals can vary from 0 to f 200V. The figure 3.16 shows the most relevant actuator characteristics.

61 CHAPTER 3. WIND TUNNEL TESTS Figure 3.14: Schematic view of the piezoelectric sensor (small patch) and actuators (big patches) placement in the active spar (gray beam) wing. Figure 3.15: Lead-zirconate-titanate piezoelectric sensors.

62 CHAPTER 3. WIND TUNNEL TESTS Peak-to-Peak Force, F [Ibs 1 Figure 3.16: ACX Quickpack 40W actuator characteristics.

63 CHAPTER 3. WIND TUNNEL TESTS In figure 3.16, the first plot shows the peak-to-peak strain versus excitation voltage curve, and the second shows the relation between peak-to-peak strain and peak-to- peak force, for each excitation voltage. For instance, an actuator excitation voltage of 140V can induce a 350~6 of strain at the actuator if it is not constrained, making the resulting strain value conditioned to the maximum force that the actuator can transmit, which is 651bs in this case The Controller The digital control module used was a dspace MicroAutoBox 1401/1501, shown in the figure The DS1401 base board is based on the PowerPC 603e processor that forms the main processing unit of the MicroAutoBox and weights The DS1401 also includes a high-speed serial interface with a 4-Mbyte-communication memory. The DS1501 1/0 board integrates the following components: - high performance Analog-Digital and Digital-Analog Converters units (AID and DIA); - a digital 1/0 subsystem based on the Motorola micro controller, not used for this work; - a Controller Area Network subsystem of the 1/0 board, based on the Siemens SAB 80C167 micro controller, primarily used for working with expansion boxes in asynchronous working modes, not used for this work. The DS1401 base board and the DS1501 1/0 board are connected via an internal bus. Both boards need a single DC power supply of 12V. Communication with the host PC is done through a dedicated DS815 PCMCIA board. This allows not only the uploading of programs to the DS1401, but also the real-time monitoring and recording variables and their consequent download [43].

64 CHAPTER 3. WIND TUNNEL TESTS Figure 3.17: Photograph of the dspace MicroAutoBox l4ol/l5ol. The design of real-time applications to the DS1401 was made using MATLABs Simulink and a dedicated toolbox provided by dspace. This toolbox permits access to internal DS1401/1501 board components, such as A/D, D/A and memory units. After building the Simulink model, the real-time Workshop is used to generate C language source code for the real-time application and start the real-time Interface. The real-time Interface automatically connects MATLAB, Simulink and the real-time Workshop with DS1401/1501 real-time system. The D/A and A/D converters timing and synchronization is selected and guaranteed by the real-time Workshop step-time (limited by the converter settling times already stated above), and thus providing the seamless implementation of the Simulink model even using additional toolboxes and continuous timeldiscrete-time hybrid systems. The dspace7s ControlDesk software enables, through a fully graphical interface, not only to visualize, record and transfer data from the DS1401 board to the host PC (both in binary or.mat format files), but also, real-time parameter tuning, like offset and gain values, and data capturing and recording parameters [44].

65 CHAPTER 3. WIND TUNNEL TESTS Figure 3.18: Photograph of the SA-10 power amplifier Electronic Equipment In terms of electronic equipment, there were used the following four components: an amplifier, a signal conditioning circuit, a power supply and voltage regulators. The used amplifier is a modified Sensor Technology SA-10 High-Voltage Power Amplifier, shown in the figure 3.18, that can be used as two individual ground- referenced amplifiers each one with a 15 times gain. The total input voltage is lim- ited to approximately -/ + 9V setting the output to a maximum voltage swing of -/ + 140V. The maximum output current per channel is internally limited to 50mA. The amplifier slew rate is 3.8V/s. One of the foremost advantages of the modified amplifier is its weight of only 100g and the small dimensions of 1" x 3" x 5/', that allows its usage in the RPV. The need of a signal conditioning arises from the fact that the output voltage signals from the sensors (-2.5V to +2.5V) are incompatible with the input voltage signals from the DS1501 (OV to +5V). Also, the voltage signals from DS1501 (OV to +4.5V) need to be modified to fit SA-10 amplifier voltage input signals (-1 + 9V), in order to achieve the amplifier's maximum output voltage to feed the actuators (-/ + 140V). On the other hand, the signal before the SA-10 amplifier must have the DC component removed. The option of trying to use the maximum output voltage

66 CHAPTER 3. WIND TUNNEL TESTS Figure 3.19: Photograph of the signal conditioning circuit and SA-10 amplifier box. of the actuators is justified because it is required the maximum output energy with the minimum controller gain, in order to achieve high strain rates to control higher frequency vibrations. The build conditioning circuit was based in the one designed by [45] and was mounted before DS1501, as shown in figure Since it was decided to limit the operations performed by hardware components (although they are faster and more efficient than software components) because of volume and weight restrictions of the UAV, the main responsibility of performing filtering and control operations was left to software modules running on the PowerPC 603e processor of the dspace. The main piece in the signal conditioning circuit is the 741 Operational Amplifier. It needs two power signals (+Vcc and -Vee). Normally +Vcc = Vee and lies between 5 and 15V. The value of +Vcc = Vee = 12V was chosen for this work, to be compatible with dspace and SA-10 power supply inputs. Although a lot of testing in the wind tunnel was conducted with the energy supply provided by the electrical network, the final wind tunnel tests were performed with two batteries in serial connection, shown in figure The positive terminal of the first battery and the negative terminal of the second battery provide the reference

67 CHAPTER 3. WIND TUNNEL TESTS Figure 3.20: Photograph of the 16.8V (left) and 14.4V (right) batteries. signal for all equipments. The first battery is a PowerfLite 14.4V, 2000mAh lithiumion pack. It is used to provide voltage and current to the -12V voltage regulator, in order to feed the -12V amplifier and signal conditioning circuit inputs. Since it feeds the negative part of the circuits, it needs to provide low output current values (O.lA), making the endurance rise to over 2 hours and 30 minutes in continuous working mode. This pack weighs 205g. To feed the +12V input of the amplifier, signal conditioning circuit and DS1401 (the most demanding component in terms of current), a 16.8V NEXcell, 2200mAh and 428g pack is used, connected to a voltage regulator. To feed these components, this battery has to continuously provide 1.2A, and more than 2.05A peak, due to the transient input peak in the ds1401 start-up. Therefore, this is the critical electrical component in terms of endurance. Since a stable voltage power supply is a fundamental requirement of all components, it was decided to include voltage regulators in the hardware components, between the two batteries and the signal conditioning circuit. The selected components were the MC7912 and LM317K. The MC7912 component provides, on its output gate, a stable voltage of -12V. It allows current loads up to 1A and is manufactured by Sarnsung GmbH. The LM317K provides, on its output gate, a stable voltage of +12V.

68 CHAPTER 3. WIND TUNNEL TESTS 5 7 It works with current loads up to 3A and is manufactured by Thomson GmbH. As this component deals with relatively high values of power, it was necessary do adapt one heat exchanger, as shown in figure Tests Objectives and Procedures The general objective of the wind tunnel tests is to demonstrate the advantage of us- ing piezoelectric sensors and actuators to actively control wing structure vibrations. First, the tests were made with the wing in free vibration, i.e., without the piezo- electric~ control. Then, similar tests were made but with active piezoelectric control working. It was verified that, with the active control working, the structure damping has increased significantly. In these tests the goal is to reduce the wing vibrations amplitude (and corresponding stress levels) and increase its damping. Second, wing flutter tests were performed. In these tests the wind speed was sussevely increased, until the interaction between aerodynamic, inertial and structural forces became un- stable. Some of these tests were destructive. In both wings, as explained in Section 3.1.2, the piezoelectric actuators were mounted near the wing root. This piezoelectric placement configuration generates, when the piezoelectric control is working, a bend- ing moment that counterbalance the wing vibrations along its span (see figure 3.10). In other words, the active wing concept used consists in controling flutter vibrations by actuating the wings first bending mode. The goal is to prevent the interaction between the first bending mode and the first torsion mode, preventing the wing flutter and thus increasing the wing resonance frequency by increasing the damping, when compared with the passive wing. The entire hardware system used for wind tunnel and flight test is shown in figure This figure highlights the connections between all hardware blocks, which were

69 CHAPTER 3. WIND TUNNEL TESTS 58 developed and assembled in a modular way. These connections are detailed in the figures legend and include power, signal and digital information flux. All blocks were disposed by functional levels (batteries, voltage regulators, sensors, signal conditioning, control, amplifier and actuators), that were exposed in more detail in Section 3.1. The most important part of this scheme are the modules between sensors and actuators. Thus, monitoring the shape of the wing in real-time, the piezoelectric sensors measurements provide real-time information, required for the feedback wing deformation control, which will determine the operation of the piezoelectric actuators. In other words, the piezoelectric sensors produce a variable f 2.5V electrical signal that is conditioned and sent to the data acquisition system (analog-to-digital converter) of the controller. Then, the controller output passes thought a digital-to-analog converter, and the analog signal is once more conditioned and sent to high-voltage power amplifier. The output signal of the amplifier is used to drive the piezoelectric actuators. In terms of control model, the classical PID controller would be a possible solution to the problem [46]. Though, as the sensors and actuators do not react to position and do not follow a reference position (they react to strain changes) the tuning of the PID parameters would not be easy, assuming that they would change with speed. Prior studies [47, 481 demonstrated that simple proportional control law, with prefilter, produced very good results in real-time conditions. Taking all these facts in consideration, the control approach selected was Proportional control, and it was developed with the goal of working in flight. The actuators will react to the provided sensors signal multiplied by a constant. This way, the conceptual approach followed in order to design the control model is presented in the figure The plant to be controlled is the active wing (active skin or active spar). The strain rate signal, measured by the sensors, is applied to the controller and the loop is closed by the

70 CHAPTER 3. WIND TUNNEL TESTS Figure 3.21: Scheme of the complete hardware system for the wind tunnel and flight tests.

71 CHAPTER 3. WIND TUNNEL TESTS - - Amplifier Pieraefectric Active Actuat@r Wing Sensar I, Piezoelectric Figure 3.22: Block diagram of conceptual active wing control model. actuators, which induce a strain, limited by the available force, to the portion of skin/spar under their location, forcing the skinlspar to move. The control model used for wind tunnel tests was designed in Simulink and is shown in figure The first block of the control model is the A/D converter output, which input is the signal of the sensor (after the signal conditioning that adds 2.5V to the output sensor signal of f 2.W). A constant value of 0.5 (5V = 1) is subtracted to this signal to remove the DC component, to obtain again a f 2.5V signal. The signal is then filtered and multiplied by two gain values (one for large adjustments and one for small adjustments) and is separated in two: the positive component goes to D/A channel 1 and the negative part, after being inverted, goes to D/A channel 2. The saved signals are controller input without filtering, controller input after filtering and controller output. As presented in Section 2.3.1, the system identification using finite element analysis revealed that the most important vibration mode for the first flutter (occurring at 44.87mls) of the active skin wing was the seventh mode at 71.9Hz. Also, since

72 CHAPTER 3. WIND TUNNEL TESTS V o%etl r mi n Come Gain I3 X Product DAC DAC-TYPEI-Mi-C1 v D%et M Offset-Sensor Fine Galn DAC Figure 3.23: Diagram representing the control model implemented for wind tunnel tests. the controller is working in order to control bending vibrations, the first mode at 16.4Hx (first bending mode) is very important. For the active spar wing (see Section 2.3.2), the revealed most important mode for the flutter occurrence was the third mode (i.e., first torsion mode) at 46.7Hx. Additionally, the first mode at 19.8Hz (first bending mode) and the second mode at 32.6Hz (second bending mode) are also very important, since the controller is working in order to control bending vibrations. Thus, the system bandwidth is a frequency range between 16.4 and 71.9Hx for the active skin wing, and between 19.8 and 46.7Hx for the active spar wing. According to Nyquist sampling theorem, the sample rate must be at least higher than twice the highest frequency of the signal to be sampled to avoid the aliasing error. Empirical considerations recommend that one should use a sampling rate five times the system bandwidth upper limit. Thus, in this work a sample rate of looohx (0.001s step time) was used to have enough samples and to allow a good frequency discrimination. This value was considered not high enough to generate over-sampling errors, and it did not trouble the AID and D/A converters. In terms of filtering, its main objective is to cut the high frequency noise. After some tests, it was decided that the filter to be

73 CHAPTER 3. WIND TUNNEL TESTS 62 used would be a second order filter with a cutting frequency of 45Hx for the active spar wing, and a second order filter with 85Hx of cutting frequency for the active skin wing. Note that, for the active spar wing, the first bending mode has a frequency of 19.8Hx and the controller is acting on this mode shape, since the filter is not cutting this frequency. The second order filter with the cutting frequency of 45Hx also allows that the wing first torsion mode is considered, thus the flutter can be studied. The same points were considered in the active skin wing: the second order filter with the cutting frequency of 85Hx allows the controller to work in the first bending mode (at 16.4Hx) and in the main frequency related with he first flutter mode (at 71.9I-l~). The wings were tested at the speeds between 10 and 37.5mls. These tests were made with the characteristic residual turbulence of the wind tunnel (2%) and a 5Hx vibration caused by the support model empennages flexibility. With the angle of attack of the airplane nearly calibrated for each speed, the signals of displacement of the wing were recorded for each flying condition. With the displacement values given by the sensor signal, the average and maximum displacements of the passive and active wings were used to calculate the improvements of this technology, shown in Sections 3.3 and 3.4. It is important to notice that the recorded signals correspond to the sensors output voltage, directly proportional to the wing vertical displacement. During tests, the values of proportional gain were tuned watching the controllers output signal and the sensors signal in ControlDesk experiment. The maximum gain that did not saturated the actuators, and managed to reduce the amplitude of controllers input signal was the one chosen to each specific speed.

74 CHAPTER 3. WIND TUNNEL TESTS Adaptive Skin Wing Tests Vibration Tests The RPV with the active skin wing was tested in the wind tunnel at several operating speeds, and with the controller in open and closed loop in order to quantify the benefits of this technology in comparison to a normal passive wing. The signals of displacement of the wing were recorded for each flying condition. After this, the results of these displacement signals (average and maximum values) with control ON (active wing) and control OFF (passive wing) were studied. For this wing, the results of the displacement signals with control ON and control OFF are shown in the table 3.2. Using the results shown in this table, the improvements of using the active wing were calculated. These improvements are shown in table 3.3. Note that both the average displacement and the maximum displacement are generally lower with the active wing than with the passive one, making the active wing an improved wing version compared with the passive wing, in the majority of flight conditions. The results shown in table 3.2 can also be graphically seen in figures 3.24 and 3.25, where the improvements of the active skin wing can be verified more explicitly Damping Analysis In order to have a deeper understanding of the aeroelastic characteristics of the wing with the controller on, the study of damping at several key frequencies was preformed. The damping influence was studied with two objectives: to quantify the amount of damping imposed by the controller in vibrations induced by the tail and to quantify the damping in the vibration modes associated with the wing flutter. All damping

75 CHAPTER 3. WIND TUNNEL TESTS Table 3.2: Average and maximum displacement values for the passive and active skin wing configurations. PASSIVE WING ACTIVE WING Speed [m/s] Average dis- Maximum Average dis- Maximum placement displacement placement displacement Table 3.3: Displacements improvements of the active skin wing compared with the passive skin wing. Speed [m/s] Improvement of the av- Improvement of the maxerage displacement /%I imum displacement f%l OO (no improvement) (noimprovement)

76 CHAPTER 3. WIND TUNNEL TESTS El Passive wing WActive wing Speed [mk] Figure 3.24: Average displacements of both passive and active skin wing configurations, in the wind tunnel-tests. Passke wing Active wing Speed [mk] Figure 3.25: Maximum displacements of both passive and active skin wing cordigurations, in the wind tunnel tests.

77 CHAPTER 3. WIND TUNNEL TESTS Figure 3.26: Damping calculation scheme. values were calculated using the FFT of the motion of the wing with and without control, throughout the speed range between 10mls and 37.5mls. The aerodynamic shape of the tail in conjunction with the vibration modes of the aircraft fuselage generate a steady vibration at around 5Hx, which shifts to almost lohx as the speed increases. After validating this mode shapes, the signals of the sensors of the wing were transformed to frequency versus amplitude in order to calculate the damping of the wing at several speeds, in the active and passive mode. Like can be seen in figure 3.26, the damping value is the ratio between the difference of the two half-power points amplitude and the natural frequency point amplitude [49]. After analyzing the FFT peak of the vibration imposed by the tail to the wing, the damping values were calculated and the results are shown in table 3.4 and figure The tail vibration peak started at 5Hx at 10mls and shifted to lohx at 30mls. The results displayed in figure 3.27 show that the damping values of the wing in the active mode are higher than in the passive mode in almost all the flying

78 CHAPTER 3. WIND TUNNEL TESTS Table 3.4: Damping of the active skin wing due to tail vibration, tested at 10, 15, 20, 25, 35 and 37.5mls. Speed [m/s] Passive Wing Damping Active Wing Damping m Passive wing -.c 0.8 E g 0." -A- Active wing Speed [m/s] Figure 3.27: Damping curves of both passive and active skin wing configurations, due to the tail vibration.

79 CHAPTER 3. WIND TUNNEL TESTS 68 conditions. These results demonstrate that the controller is improving the aeroelastic characteristics of the wing and decreasing forced vibrations induced to the structure Flutter Analysis The flutter tests were conducted in order to verify the predicted flutter speed of the wing working in the passive mode of 44.87mls (calculated in Section 2.3. I), and then to investigate the maximum attainable speed increase without fluttering the wing with the controller working. The flutter speeds (with controller on and off) were determined experimentally by calculating the damping of the vibration modes that will induce the first flutter to appear (mainly the seventh mode at 71.9Hx). Using the FFT analysis on the controllers input signal amplitude of the wing, the damping values of these modes were obtained, as shown in table 3.5. Then, a plot of damping versus speed was designed and a second order polynomial was passed through the points in order to find the zero damping speed, which corresponds to the flutter speed, as shown in figure This method of finding the flutter speed was validated by application in previous works [48]. Table 3.5: Damping of the wing for the seventh natural mode, tested at 10, 20, 25, 30 and 37.5rnls. speed [m/s] Passive Wing Damping Active Wing Damping

80 CHAPTER 3. WIND TUNNEL TESTS 13) OD n P as sive wing 0 a35 n Activevdng Poly. passive twg).= OD25 E 0.02 a OD1 5 a -Poly. (Actke wing$ 0.01 OD Speed [dsj Figure 3.28: Curves of the wing damping for the seventh natural mode at 71.9Hx, and polynomial extrapolation to zero damping. Looking to the results in figure 3.28, it can be seen that the wing damping for the seventh mode is always higher in the active mode than in passive one. Then, it can be concluded that the active wing has a higher flutter speed when compared with passive wing, as can be verified by the extrapolated curve to the zero damping. The coefficient of determination R2, which represents the proportion of the variance of the damping with respect to the speed, is for the polynomial representing the passive wing, and for the active wing polynomial. Using these results, the flutter speed of the wing in the passive mode is 47.95~~1~ and in the active mode is 53.99rn/s, with variations defined by the previous referred R2 values. This represents an increase in flutter speed of approximately 12.59%, meaning that with a given wing and a control system like this, one can fly 12.59% faster, without structural reinforcements of the structure and without suffering the aeroelastic effects of flutter. Finally, comparing

81 CHAPTER 3. WIND TUNNEL TESTS 70 the flutter speed of 44.87mls calculated in computational analysis, in Section 2.3, with the experimental value of 47.95m/s, there is an 6.86% error comparing the experiment a1 value with the computational one. 3.4 Adaptive Spar Wing Tests Vibration Tests As performed with the active skin wing, the RPV with the active spar wing was tested in the wind tunnel at several operating speeds, and with the controller in open and closed loop. The results of the displacement signals, with control ON (active wing) and control OFF (passive wing), were recorded for each flying condition. These results are shown in the table 3.6. Using the results shown in this table, the improvements of using the active wing were calculated. These improvements are shown in table 3.7. Like concluded with the active skin wing results, both the average displacement and the maximum displacement are generally lower with the active spar wing than with the passive one, making the active wing an improved wing versidn compared with the passive wing, in the majority of flight conditions. The results shown in table 3.6 can also be graphically seen in figures 3.29 and These tests were very helpful because the wind tunnel model was in a configuration in which the tail was very flexible, in order to force vibrations on the wing besides the aerodynamic induced vibrations. Then, the tests were repeated but without the flexible tail, in order to simulate the flying version of the RPV. Similar analysis was performed with this RPV configuration, and the obtained results are shown in tables 3.8, 3.9 and figures 3.31, Note that the improvements are generally much better in this condition than in the configuration that includes the flexible tail. These

82 CHAPTER 3. WIND TUNNEL TESTS 71 tests proved that the proposed approach for active aeroelastic control significantly decreases both the maximum and average displacements that a wing sustains during flight. This way, the fatigue life will increase leading to a better structure. Table 3.6: Average and maximum displacement values for the passive and active spar wing configurations. PASSIVE WING ACTIVE WING Speed [m/s] Average dis- Maximum Average dis- Maximum placement displacement placement displacement Table 3.7: Displacements improvements of the active spar wing compared with the passive spar wing. Speed [m/s] Improvement of the av- Improvement of the maxerage displacement [%I imum displacement [%] (no improvement)

83 CHAPTER 3. WIND TUNNEL TESTS Passive wing RActive wing Speed [mls] Figure 3.29: Average displacements of both passive and active spar wing configurations, in the wind tunnel tests. Table 3.8: Average and maximum displacement values for the passive and active spar wing configurations, using the RPV without the flexible tail. PASSIVE WING ACTIVE WING Speed (m/sl -. - Average - dis- Maximum Average dis- Maximum placement displacement placement displacement

84 CHAPTER 3. WIND TUNNEL TESTS ' Passive wing Activ e wing Speed [mk] Figure 3.30: Maximum displacements of both passive and active spar wing configurations, in the wind tunnel tests. Table 3.9: Displacements improvements of the active wing compared with the passive wing, using the RPV without the flexible tail. Speed [m/s] Improvement of the av- Improvement of the maxerage displacement [%I imum displacement [%I

85 CHAPTER 3. WIND TUNNEL TESTS Passive wing El Active wing Speed [mfs] Figure 3.31: Average displacements of both passive and active spar wing configurations, in 1 ;he wind tunnel tests (RPV without tail) I P L E 0.1 E - m TT E X % 0.02 Passive wing Active wing Speed [mls] Figure 3.32:- Maximum displacements of both passive and active spar wing configurations, in the wind tunnel tests (RPV without tail).

86 CHAPTER 3. WIND TUNNEL TESTS Damping Analysis Similarly as done with the active skin analysis, a study of damping at several key frequencies was preformed. The damping influence was studied with the same two objectives: to quantify the amount of damping imposed by the controller in vibrations induced by the tail and to quantify the damping in the vibration modes associated with the wing flutter. All damping values were calculated using the FFT of the motion of the wing with and without control, throughout the speed range between 15mls and 30mls. After analyzing the FFT peak of the vibration imposed by the tail to the wing, the damping values were calculated and the results are shown in table 3.10 and figure The tail vibration peak started at 5Hx at 15mls and shifted to 8Hx at 30mls. Analyzing the figure 3.33, one can conclude that the damping values of the wing in the active mode are always higher than in the passive mode, even though the differences at speeds higher than 25mls are not very significant. These results demonstrate, again, that the controller besides improving the aeroelastic characteristics of the wing can also decrease forced vibration induced to the structure, and is more efficient in this task at low speeds, around the cruising speed. Table 3.10: Damping of the active spar wing due to tail vibration, tested at 15, 20, 25 and 30mls. Speed [m/s] Passive Wing Damping Active Wing Damping

87 CHAPTER 3. WIND TUNNEL TESTS P 0.8 *Passive wing Active wing 0! I I I Speed [mls] Figure 3.33: Damping curves of both passive and active spar wing configurations, due to the tail vibration Flutter Analysis As stated in Section 3.3.3, the flutter tests were conducted with two objectives: to verify the predicted flutter speed of the passive wing working (calculated in Section 2.3.2), and to calculate the increase of the flutter speed related with the use of the controller (active wing). The flutter speeds (controller on an off) were determined experimentally by calculating the damping of the vibration modes that will induce flutter (first torsion mode at 46.7Hx). Using the FFT analysis on the controllers input signal amplitude of the wing, the damping values of the first torsion modes were obtained, as shown in table Then, a plot of damping versus speed was designed and a second order polynomial was passed through the points in order to find the zero damping speed (corresponding to the flutter speed), as shown in figure Note that the damping of the first torsion mode is the most important to

88 CHAPTER 3. WIND TUNNEL TESTS 77 determine the flutter speed, since this wing has a conventional configuration (i.e., with a main beam), thus, it flutters conventionally. The results in figure 3.34 show that the wing damping for the first torsion mode is always higher in the active mode than in passive one. Then, one can state that the active wing has a higher flutter speed when compared with passive wing, as can be verified by the extrapolated curve to the zero damping. The coefficient of determination R2, which represents the proportion of the variance of the damping with respect to the speed, is for the polynomial representing the passive wing, and for the active wing polynomial. With these results, the flutter speed of the wing in the passive mode is 44.38rnls and in the active mode is 53.71m/s, which represents an increase in flutter speed of 21.02%, with speed variations defined by the previous referred R2 values. This means that with a given wing and a control system like this, you can fly approximately 21.02% faster without structural reinforcements of the structure and without suffering the aeroelastic effects of flutter. Also, in figure 3.34 it can be seen that the difference in damping values on active and passive modes increases with the test speed, suggesting that the controller is working to suppress flutter at high speeds (over 25mls). Comparing the passive wing flutter speed calculated in Section 2.3.2, of 49.83m/s7 with the experimental value of 44.38m/s, there is an error of 12.28%.

89 CHAPTER 3. WIND TUNNEL TESTS Table 3.11: Damping of the wing for the first torsion mode, tested at 15, 20, 25, 27.5 and 30mls. Speed [m/s] Passive Wing Damping Active Wing Damping I O " '?A\ R* = =% ' C , > Speed [mls] D.05 P A PassiVewing ActiveLving ---- Paty'. (Passiw wing) - Pa&. (Ac t'm wing) Figure 3.34: Curves of the wing damping for the first torsion mode, and extrapolation polynomial to zero damping.

90 Chapter 4 Flight Tests 4.1 Experimental Apparatus The hardware involved in the flight tests is similar as the described in the Chapter 3. However there are some differences, as is next described. The flight tests were performed in the Sintra Air Base takeoff runway, Portugal. In terms of tests articles, the only difference is the RPV platform and the required engine. A new and improved RPV platform was used, which is described in Section The flight tests were performed using the adaptive spar wing only, since is lighter than the adaptive skin wing. The same piezoelectric sensors and actuators were used. The used digital controller was also the same as described in Chapter 3, although the control law has some differences. In terms of electronic equipment, some additional components were used. Therefore, the next Sections include the description of the following hardware components: the RPV and the additional electronic equipment.

91 CHAPTER 4. FLIGHT TESTS TheRPV The RPV fuselage is lighter than the one used for the wind tunnel tests, in order to have more payload capabilities. This fuselage is less stiff than the other. It was constructed mostly using balsa wood. Also, carbon fibre was used in the places where more stiffness was necessary. The used RPV engine has a power of 1.864kW. The tables 4.1,4.2,4.3 and 4.4 present the RPV main characteristics in terms of geometry, aerodynamics and performance. The figure 4.1 shows the final RPV model ready to start a flight test. Note that the maximum speed at sea level is very high. However, and fortunately, it is smaller than the flutter speed of the passive wing. On the other hand, the stall speed in clean configuration is higher than 15 m/s, which was one wind tunnel testing speed. This speed was tested only to check the performance of the control system. Table 4.1 : RPV external dimensions. Parameter in S.I. Units Main Wing Span 2.4 m Main Wing Chord (Root) 0.33 m Main Wing Chord (Tip) 0.33 m Main Wing Aspect Ratio Horizontal Stabilizer Span 0.72 m Horizontal Stabilizer Chord (Root) 0.18 m Horizontal Stabilizer Chord (Tip) 0.18 m Horizontal Stabilizer Aspect Ratio 4 Vertical Stabilizer Span 0.2 m Vertical Stabilizer Chord (Root) m Vertical Stabilizer Chord (Tip) m Vertical Stabilizer Aspect Ratio Wheel Track m Wheel Base m Propeller Diameter 0.38 m

92 CHAPTER 4. FLIGHT TESTS Table 4.2: RPV areas. Parameter b21 Main Wing Area Flaps Area Ailerons Area Horizontal Stabilizer Area Elevator Area Vertical Stabilizer Area Rudder Area Table 4.3: RPV weights and loadings. Parameter in S.I. Units Maximum Takeoff Weight 10 Kg Maximum Wing Loading kg/m2 Maximum Power Loading kglwatts Table 4.4: RPV performance data. Parameter b/si Maximum Level Speed (See Level) Cruise Speed (See Level, 75% Power) 38.6 Stall Speed Clean Stall Speed (45" Flaps) Maximum Rate of Climb 8.44

93 CHAPTER 4. FLIGHT TESTS Figure 4.1: Picture of the RPV model ready to the flight tests Additional Electronic Equipment In terms of electronic equipment, the same components as described in Section were used, i.e., the controller, the amplifier, the signal conditioning circuit, the two power supply batteries and voltage regulators. However, during the flight tests it is important to have the real time information about the RPV airspeed. Thus, a telemetry system with bidirectional data link was installed in the RPV, and the speed information was available during the entire flight envelope. This system has an airborne station and a ground one. The components of the airborne station are the following (see figure 4.2) : - a microcomputer, which performs the signal conditioning, and radio frequency modulation operations; it is connected to the battery pack, pitot's transducer, air station transceiver and onboard radio control receiver; - a 7.2V, 1250mAh, 6 cell Ni-Cd battery pack, that supplies the power to the airborne station; - a pitot sensor (total pressure and static pressure inputs), with a transducer to convert pressure in electronic signal, and a pulse position modulation modulator;

94 CHAPTER 4. FLIGHT TESTS Figure 4.2: Photograph of the telemetry airbone station components. - a transceiver with antenna, which sends radio signals to the ground station; - a radio-control receiver with 9 channels, which has an independent battery pack. The ground station components are (see figure 4.3): - a transceiver that receives the radio signals from the aircraft; it is connected to the PC by the USB slot (power) and RS232 serial adapter (data); - a radio-control unit, used by the pilot to fly the RPV; the ground system doesn't work if this unit is turned off. The software interface between the user and the telemetry system is the Jet-tronic I1 for Windows. This program has several applications, such as real-time changing parameters, and showing its real-time values in the data display. The figure 4.4

95 CHAPTER 4. FLIGHT TESTS Figure 4.3: Photograph of the telemetry ground station components. illustrates the data display window. It shows throttle and trim controls position, fuel pump voltage, engine RPM and EGT (Exhaust Gas Temperature), airspeed and battery voltage. The last two functions were the only ones used for the flight tests. The figure 4.5 shows the RPV fusclage containing all the electronic components necessary to perform the flight tests. 4.2 Tests Objectives and Procedures The objective of the flight tests is to prove that the piezoelectrics vibration control technology works not only in a controlled environment (like the wind tunnel), but also in a real environment, and it is portable enough to fly in a small airplane like this one. During the flight, the control was automatically switched on and off every 12sec, in order to compare the differences between active and passive wing configurations, and

96 CHAPTER 4. FLIGHT TESTS 85 Figure 4.4: Picture of the real-time data display. Figure 4.5: Photograph of the RPV fuselage containing all the flight tests equipment.